AI Briefing#world-models
The latest AI and developer news about #world-models, with the original source and a short summary.
The latest AI and developer news about #world-models, with the original source and a short summary.
The model couples a Qwen3.5-9B language model with a 29B MMDiT diffusion backbone to predict video outcomes of actions.
The open-source framework provides a unified catalog and execution pipeline to separate execution readiness from evaluative claims across video, 3D, and 4D systems.
The open-source framework pairs automatic data generation with benchmarks showing VLA models generalize poorly while frontier agents solve 30/30 stacking tasks.
The open-source Cosmos Framework supports training and simulation for autonomous vehicles, robotics, and surgical workflows, while NIM enables sub-second live stream alerts.
The model surpasses leading open-weight models on MindCube spatial reasoning and outperforms other world models on FlowBench motion prediction.
The expanded platform includes five cloud-native data blueprints, support for fine-tuning open models like GR00T and QWEN, and Cosmos 3 Nano variants that enable up to 2x faster inference.
Gemini 4 Argon processed 25 trillion tokens in 72 hours, tripling Gemini 3.0 usage, while Pichai confirms compute constraints are driving efficiency breakthroughs.
Tencent ARC Releases WorldCrafter: Open-Source Video World Model with Implicit 3D Memory
The model uses a new dataset of 20,774 clips and a Motion Adapter to ensure view-consistent video generation.
A 201M-parameter SemanTok AR model matches or outperforms a VideoFlexTok AR model 3.4x its size.
ByteDance is set to launch a cloud-rendering-based real-time spatial video model next month under the supervision of Zhang Yiming.
World Labs co-founders discussed the world model Atlas and the future of spatial intelligence on the a16z podcast.
Yann LeCun, Demis Hassabis, and Fei-Fei Li have identified decision-centric World Models as the next stage of AI.
World Labs has unveiled 'Atlas', an integrated world model that generates camera-controllable 1440p HD video from text, image, and 3D data inputs.
World Labs has released Atlas, a spatial intelligence-specialized omni world model that integrates text, image, and 3D processing.
BabyZWM, trained solely on a single child's visual experience, matched SOTA on multiple visual-cognitive tasks.
HY-World 2.0 has been released, creating editable 3D worlds from text, images, and video.
Waypoint-1.5 extends real-time world models for consumer GPUs with 720p/60FPS and 360p tiers.